MCP & Memory Intern

DataRobot•Boston, MA

About The Position

DataRobot delivers AI that maximizes impact and minimizes business risk. Our platform and applications integrate into core business processes so teams can develop, deliver, and govern AI at scale. DataRobot empowers practitioners to deliver predictive and generative AI, and enables leaders to secure their AI assets. Organizations worldwide rely on DataRobot for AI that makes sense for their business — today and in the future. This effort is about making AI agents more useful in real work: giving them the right tools, the right context, and a memory that holds up across a conversation and across sessions. The focus is Model Context Protocol (MCP) and agent memory, how an agent discovers tools, calls them safely, keeps track of what it has learned, and stays helpful without someone watching every step. The goal is practical value for people using agents day to day, with clean implementations that others can build on. Summary: This is a hands-on internship for someone curious about AI agents who wants to learn by building. You do not need to already be an expert. What matters is that you like figuring things out, you move without waiting to be told every next step, and you care about unblocking the people around you. Most of the work sits on MCP and memory: connecting agents to tools, shaping how context is passed in, and designing how an agent remembers what is useful and forgets what is not. You will write code, try ideas, see what breaks, and ship improvements in a fast-moving environment. A good attitude and the willingness to learn matter more than a perfect resume.

Requirements

Comfort writing and reading Python, and a willingness to get better at it quickly.
Curiosity about how AI agents work: tools, prompts, context windows, and why an agent succeeds or fails on a real task.
Interest in MCP — how servers expose tools, how clients call them, and how that contract stays simple and reliable.
Interest in memory for agents: what to store, what to retrieve, and how that changes what the agent can do next.
Ability to take a vague problem, try a small version, and show what you learned.
Clear communication: you say when you are stuck, you share what you tried, and you help other people move faster.
Good engineering habits you are willing to practice: readable code, small tests, and notes so the next person is not lost.
You take feedback on board. When someone points out a better approach, a mistake, or a simpler path, use it and adjust.

Nice To Haves

You have built or used an MCP server or client, even a small one.
You have played with agent frameworks or tool calling.
You have thought about memory, caching, or conversation state in any product, not only in AI.
You enjoy fast environments and would rather try something today than wait for a perfect plan.
You like helping teammates get unblocked, even when the task is not yours.

Responsibilities

Building and improving MCP servers and clients: tool definitions, inputs and outputs, error handling, and examples other people can run.
Experimenting with agent memory: what gets saved, how it is retrieved, and how that changes task quality.
Wiring tools and memory into an agent loop and watching where it gets confused, slow, or wrong.
Making small, clear changes that unblock teammates - docs, fixtures, smoke tests, and fixes that remove friction.
Turning a rough idea into something that runs, then tightening it based on what you see.
Pairing with people who know the system well, then owning the next slice yourself.